The EU AI Act and AI Phone Calls: What Changes on 2 August 2026

From 2 August 2026, a caller in the EU has a legal right to know when the voice on the other end of the line is an AI. Article 50 of the EU AI Act — the transparency chapter — becomes applicable on that date, and it covers exactly the kind of system more and more businesses now use on their phones: AI receptionists, voice agents, phone assistants. Here is what changes, who it applies to (including UK businesses), and how to be ready with an afternoon's work.

In short: from 2 August 2026, AI systems designed to interact with natural persons must inform those persons that they are dealing with an AI, unless that is obvious from the circumstances (Article 50, EU AI Act). On a phone call it is rarely obvious, so a clear disclosure at the start of the call is the safe answer. The obligation follows the market, not the company seat: it applies to providers and deployers serving the EU market wherever they are based — including UK and Swiss businesses with EU customers. Penalties for transparency violations run up to €15M or 3% of global turnover.

*(This article is general information, not legal advice — check your specifics with a qualified adviser.)*

What does Article 50 actually require?

The EU AI Act (Regulation (EU) 2024/1689) regulates AI by risk category, but its transparency obligations in Article 50 apply across the board. For AI on the phone, two duties matter:

Disclosure of AI interaction. AI systems intended to interact directly with natural persons must be designed so that the person is informed they are interacting with an AI — unless this is obvious to a reasonably well-informed, observant person given the circumstances. Modern AI voices are good enough that "obvious" is not something you should rely on in a phone call. A plain opening line does the job: "Hello, you're speaking with the AI assistant of Smith & Co."

Marking synthetic audio. AI-generated audio content must be identifiable as artificially generated. For a voice agent, the practical route is the same disclosure at the start of the call.

Three details are easy to get wrong:

• The disclosure belongs at the start of the interaction. A line buried in your terms and conditions or privacy policy does not inform a caller mid-conversation. Say it in the greeting.
• Risk category is irrelevant here. Even a minimal-risk appointment-booking assistant must disclose. There is no carve-out for simple use cases.
• Penalties are real. Transparency violations carry fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher, with proportionality for SMEs.

Who is affected?

The AI Act applies to providers (companies that build and supply AI systems) and deployers (companies that use them) — and its reach follows the EU market, not the corporate registry:

• Businesses established in the EU that use an AI phone assistant: covered, from 2 August 2026.
• Businesses outside the EU whose AI system is placed on the EU market or whose output is used in the EU: also covered. A UK or Swiss company whose AI receptionist answers calls from EU customers falls within scope regardless of where it is incorporated.
• Your AI provider carries duties too, but as the deployer you are the one whose phone line greets the caller. Choosing a provider that has disclosure built in is the simplest way to discharge your side.

What about the UK?

Post-Brexit, the EU AI Act does not apply domestically in the UK — there is currently no UK statute that mirrors Article 50. But two things mean UK businesses should not file this under "not my problem":

Serving EU customers puts you in scope. The extraterritorial reach described above applies to UK companies exactly as it does to any other non-EU company. If your AI receptionist takes calls from customers in Ireland, France or Germany, plan for Article 50 compliance by August 2026.

UK GDPR already expects transparency. The ICO's transparency principle requires that people understand how their personal data is processed — and an AI answering, transcribing and summarising calls is processing personal data. Telling callers up front that they are speaking with an AI is the natural way to honour that, and it is what the ICO's fairness and transparency guidance points towards. For the full UK picture — lawful basis, PECR, DPAs and retention — see our guide Is an AI receptionist legal in the UK?

In short: for UK businesses, disclosure is legally required the moment EU callers are involved, and best practice even when they are not.

A practical checklist

Three steps cover the essentials, and none of them takes longer than an afternoon:

1. Fix the greeting. Make sure your AI assistant identifies itself as an AI at the start of every call. Two seconds of call time, and the single most important compliance step. It also simply works better: callers who know what they are talking to state their business more directly and are not left feeling tricked.

2. Update your privacy notice. Add the AI phone assistant to your privacy policy: what it does (answers, transcribes, summarises), what data it touches, where that data is stored, and who the provider is. If you want to dig into the storage question, we have covered where an AI receptionist's data actually lives.

3. Vet your provider. Ask directly: is AI disclosure built in by default, or is it something you have to remember to configure? Where is data stored and processed? Is there a data processing agreement? A provider without crisp answers on disclosure will not be much help on the rest of your compliance either.

How fonea handles it

fonea's assistant discloses that it is an AI assistant at the start of calls — by default, as standard product behaviour, not as a setting you need to find. That covers the Article 50 disclosure duty and aligns with UK GDPR and EU GDPR transparency expectations at the same time.

The rest of the frame is built for European compliance too: persistent data is stored in Switzerland, AI processing runs transiently on servers in Europe, and every customer gets a data processing agreement. The assistant answers in five languages on one number, from £90/€90 per month with 120 minutes included, month-to-month with no minimum term.

Key Takeaways

• From 2 August 2026, AI systems interacting with people in the EU must disclose that they are AI (Article 50, EU AI Act) unless it is obvious — which, on a phone call, it rarely is.
• The rule follows the market, not the company seat: UK, Swiss and other non-EU businesses are in scope as soon as they serve EU customers.
• The UK has no domestic equivalent, but UK GDPR/ICO transparency principles already point the same way — and EU callers bring Article 50 into play regardless.
• Disclosure belongs in the greeting, not in the small print, and applies to every risk category.
• Three steps to be ready: fix the greeting, update the privacy notice, vet the provider.

Frequently Asked Questions

Is a note on our website enough?

No. The point of Article 50 is that the person knows they are talking to an AI during the interaction itself. The disclosure belongs in the call greeting; the website and privacy notice complement it but do not replace it.

Does this apply to us if we only serve UK customers?

The EU AI Act does not apply domestically in the UK, so with a strictly UK-only customer base Article 50 does not bind you. UK GDPR transparency principles still apply to the data processing involved, and disclosing the AI up front is the accepted way to meet them — so the practical answer is the same greeting either way.

What are the penalties for not disclosing?

Transparency violations under the AI Act carry fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher, enforced by the market surveillance authorities of EU member states, with proportionality applied for SMEs.

Does the disclosure requirement depend on how risky the AI system is?

No. The transparency obligations in Article 50 apply regardless of risk classification. A simple appointment-booking voice agent must disclose just as a more sophisticated system must.

We use a human-sounding AI voice. Can we rely on callers realising it's an AI?

No. The exception only applies where the AI nature is obvious to a reasonably well-informed person — and the whole point of modern voice AI is that it no longer sounds obviously artificial. An explicit disclosure is the only robust approach.

Sources

• EU AI Act, Regulation (EU) 2024/1689, Article 50 (transparency obligations): artificialintelligenceact.eu
• EU AI Act, Article 99 (penalties): artificialintelligenceact.eu
• European Commission — AI Act, application timeline: digital-strategy.ec.europa.eu
• UK Information Commissioner's Office (ICO) — transparency principle, *Guide to the UK GDPR*